Add 4-bit quantization and audioop-lts for Python 3.13

- Use BitsAndBytesConfig for 4-bit quantization to fit in GPU memory
- Load LoRA adapter from HuggingFace model repo
- Add audioop-lts dependency for Python 3.13 compatibility
- Gradio 6.5.1 with minimal dependency conflicts

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>

.gitattributes

@@ -0,0 +1,6 @@
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.gguf filter=lfs diff=lfs merge=lfs -text

.gitignore

@@ -0,0 +1,12 @@
+ceo-voice-lora/
+.DS_Store
+__pycache__/
+*.pyc
+*.pyo
+*.pyd
+.Python
+*.so
+*.egg
+*.egg-info/
+dist/
+build/

README.md

@@ -0,0 +1,65 @@
+---
+title: CEO AI Executive
+emoji: 🎯
+colorFrom: blue
+colorTo: purple
+sdk: gradio
+sdk_version: 6.5.1
+app_file: app.py
+pinned: false
+license: mit
+models:
+  - unsloth/qwen3-30b-a3b
+tags:
+  - chatbot
+  - lora
+  - qwen3
+  - fine-tuning
+---
+
+# CEO AI Executive 🎯
+
+An AI chatbot that responds like your CEO, trained on their blog posts and writings.
+
+## Features
+
+- 💬 Natural conversation interface
+- 🧠 Powered by fine-tuned Qwen3-30B model
+- 🎨 Clean and intuitive UI
+- ⚡ Fast response generation
+
+## How It Works
+
+This application uses:
+1. **Base Model**: Qwen3-30B (unsloth/qwen3-30b-a3b)
+2. **Fine-tuning**: LoRA adapter trained on CEO's blog posts
+3. **Interface**: Gradio chatbot UI
+
+The model has been fine-tuned to capture the CEO's:
+- Writing style and tone
+- Perspectives on business and leadership
+- Communication patterns
+- Domain expertise
+
+## Usage
+
+Simply type your question in the chat box and the AI will respond in the CEO's voice and style.
+
+### Example Questions
+
+- "What's your vision for the company?"
+- "How do you approach leadership?"
+- "What are your thoughts on innovation?"
+- "Can you share your perspective on team building?"
+- "What drives your business strategy?"
+
+## Technical Details
+
+- **Model**: Qwen3-30B with LoRA fine-tuning
+- **Framework**: Transformers, PEFT
+- **Interface**: Gradio 6.5.1
+- **Hardware**: GPU-accelerated (Hugging Face Spaces GPU)
+
+## Disclaimer
+
+This AI is trained on historical writings and represents patterns learned from the CEO's public content. Responses should not be considered as official statements or advice from the actual CEO.

app.py

@@ -0,0 +1,153 @@
+import gradio as gr
+import torch
+from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig
+from peft import PeftModel, PeftConfig
+import spaces
+
+# Model configuration
+BASE_MODEL = "unsloth/qwen3-30b-a3b"
+LORA_ADAPTER_PATH = "AI-Talent-Force/ceo-voice-lora-qwen3-30b"
+
+# Load model and tokenizer
+@spaces.GPU
+def load_model():
+    """Load the base model and apply LoRA adapter"""
+    print("Loading tokenizer...")
+    tokenizer = AutoTokenizer.from_pretrained(BASE_MODEL)
+
+    print("Loading base model...")
+    # Use 4-bit quantization to fit in GPU memory
+    quantization_config = BitsAndBytesConfig(
+        load_in_4bit=True,
+        bnb_4bit_compute_dtype=torch.bfloat16,
+        bnb_4bit_use_double_quant=True,
+        bnb_4bit_quant_type="nf4"
+    )
+
+    model = AutoModelForCausalLM.from_pretrained(
+        BASE_MODEL,
+        quantization_config=quantization_config,
+        device_map="auto",
+        trust_remote_code=True
+    )
+
+    print("Loading LoRA adapter...")
+    model = PeftModel.from_pretrained(model, LORA_ADAPTER_PATH)
+    model.eval()
+
+    print("Model loaded successfully!")
+    return model, tokenizer
+
+# Initialize model and tokenizer
+print("Initializing CEO AI Executive...")
+model, tokenizer = load_model()
+
+@spaces.GPU
+def chat_with_ceo(message, history):
+    """
+    Chat function that responds like the CEO
+    Args:
+        message: User's current message
+        history: List of previous messages [[user_msg, bot_msg], ...]
+    """
+    # Build conversation context
+    conversation = []
+    for user_msg, bot_msg in history:
+        conversation.append({"role": "user", "content": user_msg})
+        conversation.append({"role": "assistant", "content": bot_msg})
+
+    conversation.append({"role": "user", "content": message})
+
+    # Apply chat template
+    prompt = tokenizer.apply_chat_template(
+        conversation,
+        tokenize=False,
+        add_generation_prompt=True
+    )
+
+    # Tokenize
+    inputs = tokenizer(prompt, return_tensors="pt", truncate=True, max_length=4096)
+    inputs = {k: v.to(model.device) for k, v in inputs.items()}
+
+    # Generate response
+    with torch.no_grad():
+        outputs = model.generate(
+            **inputs,
+            max_new_tokens=512,
+            temperature=0.7,
+            top_p=0.9,
+            do_sample=True,
+            repetition_penalty=1.1,
+            pad_token_id=tokenizer.pad_token_id,
+            eos_token_id=tokenizer.eos_token_id
+        )
+
+    # Decode response
+    response = tokenizer.decode(outputs[0][inputs['input_ids'].shape[1]:], skip_special_tokens=True)
+    return response
+
+# Create Gradio interface
+with gr.Blocks(theme=gr.themes.Soft()) as demo:
+    gr.Markdown(
+        """
+        # 🎯 CEO AI Executive
+
+        Chat with an AI trained on your CEO's writing style and thoughts.
+        Ask questions about business strategy, leadership, technology, or any topic your CEO writes about.
+
+        **Note:** This AI responds based on patterns learned from the CEO's blog posts and writings.
+        """
+    )
+
+    chatbot = gr.Chatbot(
+        height=500,
+        label="Chat with CEO AI",
+        show_label=True,
+        avatar_images=(None, "🎯")
+    )
+
+    with gr.Row():
+        msg = gr.Textbox(
+            label="Your Message",
+            placeholder="Ask me anything...",
+            show_label=False,
+            scale=4
+        )
+        submit = gr.Button("Send", variant="primary", scale=1)
+
+    with gr.Row():
+        clear = gr.Button("Clear Chat")
+
+    gr.Examples(
+        examples=[
+            "What's your vision for the company?",
+            "How do you approach leadership?",
+            "What are your thoughts on innovation?",
+            "Can you share your perspective on team building?",
+            "What drives your business strategy?"
+        ],
+        inputs=msg,
+        label="Example Questions"
+    )
+
+    gr.Markdown(
+        """
+        ---
+        ### About This AI
+        This chatbot uses a fine-tuned Qwen3-30B language model trained on the CEO's blog posts and writings.
+        It attempts to replicate their writing style, thinking patterns, and perspectives on various topics.
+        """
+    )
+
+    # Event handlers
+    msg.submit(chat_with_ceo, inputs=[msg, chatbot], outputs=chatbot)
+    submit.click(chat_with_ceo, inputs=[msg, chatbot], outputs=chatbot)
+    clear.click(lambda: None, None, chatbot, queue=False)
+
+    # Clear message box after submission
+    msg.submit(lambda: "", None, msg)
+    submit.click(lambda: "", None, msg)
+
+if __name__ == "__main__":
+    demo.queue()
+    demo.launch()

requirements.txt

@@ -0,0 +1,9 @@
+gradio==6.5.1
+transformers>=4.50.0
+torch==2.5.1
+peft==0.18.1
+accelerate==1.2.1
+safetensors==0.4.5
+spaces==0.30.3
+bitsandbytes==0.45.0
+audioop-lts